README Excerpt
Shared vector search infrastructure for MCP servers. Provides dense and sparse embeddings, hybrid search with Reciprocal Rank Fusion, Qdrant vector storage, and supporting utilities (caching, file discovery, change detection, glossary, facts) as a reusable Python library. - **Dense embeddings** via any OpenAI-compatible API (llama.cpp, vLLM, Ollama, OpenAI, etc.)
Tools (20)
ErrorCollectorThreadSafeSQLiteStoreVECTOR_CACHE_DIRVECTOR_CACHE_MAX_ENTRIESVECTOR_CACHE_MAX_SIZE_GBVECTOR_CIRCUIT_BREAKER_RESET_SECONDSVECTOR_CIRCUIT_BREAKER_THRESHOLDVECTOR_COLLECTION_NAMEVECTOR_CONTENT_HASH_DISPLAY_LENGTHVECTOR_DENSE_WEIGHTVECTOR_EMBEDDING_BATCH_SIZEVECTOR_EMBEDDING_CONCURRENCYVECTOR_EMBEDDING_DIMVECTOR_EMBEDDING_MAX_TEXT_CHARSVECTOR_EMBEDDING_MODELVECTOR_EMBEDDING_TIMEOUTVECTOR_EMBEDDING_URLVECTOR_FILE_LOCK_TIMEOUTVECTOR_GLOBAL_VOCAB_CACHE_TTLVECTOR_MAX_FILE_SIZE_KB